Anal. Chem. 1989, 61, 861-863
a61
Ultralow Detection Limits for an Organic Dye Determined by Fluorescence Spectroscopy with Laser Diode Excitation Paul A. Johnson,Tye E. Barber, Benjamin W. Smith, and James D. Winefordner* Department of Chemistry, University of Florida, Gainesville, Florida 3261 1
Fluorescence of IR-140, a laser dye In methanol solution, Is exclted by a semiconductor laser diode. Analytical figures of merit are compared for three different instrumental conflgurations, with the dye measured In a cuvette, a liquid Jet, and a compact Instrument. The best ilmlt of detection, 46000 molecules, was achleved with a liquid Jet. Linear dynamic range was 6 orders of magnitude. The laser diode operates in the near-infrared region, resulting In low background fluorescence.
INTRODUCTION Ultraviolet-visible fluorescence is a selective and extremely sensitive spectrometric technique. Various workers have used this sensitivity to pursue the goal of detecting a single molecule. Hirschfeld detected individual antibody molecules tagged with 8&100 fluorescein molecules in a solution spread on a microscope slide (I). The minimum number of fluorescent tags necessary to detect a single molecule may eventually become a popular way of reporting a limit of detection (LOD). A very low LOD for Rhodamine 6G adsorbed onto 10-Mm silica spheres was achieved by Kirsch et al. (2). When analysis is performed in liquid solution, solvent and impurity fluorescence and Raman scatter must be filtered out, as they contribute to the background. Impressive LODs have been achieved both in static cells (3-5) and in flowing streams, such as a high-performance liquid chromatography (HPLC) eluent (6-9).
As well as detecting low concentrations, a technique approaching single molecule detection must excite fluorescence in extremely small sample volumes. The sheath flow cuvette has been used by Dovichi et al. (10) and Nguyen et al. (11) to produce probe volumes in the picoliter range. Excitation at 514 nm by an argon ion laser achieved superior detection limits and applicability to HPLC eluent. Low cost, compact lasers for excitation of atomic and molecular fluorescence are needed. Laser excited atomic fluorescence spectroscopy (LEAFS) has suffered from the complexity and expense of the lasers used. Semiconductor laser diodes have been used in this laboratory to determine rubidium in a graphite furnace by LEAFS (12), but the number of elements that can be determined in this way is limited by the range of wavelengths that diode lasers can cover. The radiation from a frequency doubled laser diode (LD) proved insufficient to determine metals in a flame or graphite furnace. Excitation of molecular fluorescence with laser diodes has been much more fruitful, because of the absence of flame or furnace emission background and the much broader excitation bands. Fluoresence from the solvent and impurities in the solvent often limit the ultimate sensitivity of techniques using UV or visible lasers for excitation (13). Near-infrared (near-IR) fluorescence avoids this limitation, since few compounds fluoresce in this wavelength region. While this also limits the compounds that can be directly determined, the low background leads to an ultralow LOD for those com0003-2700/89/0361-0881$01.50/0
pounds that do fluoresce. Raman scatter background is less than with UV-vis lasers since the scattering process is less efficient in the near-IR region and the scatter is shifted beyond the spectral range of most photomultiplier tubes. In addition, problems associated with sample photodecomposition are much reduced at longer excitation wavelengths. The synthesis of near-IR fluorescent labeling reagents for fluoroimmunoassay and automated DNA sequencing would capitalize on this situation and open up new prospeds in the biochemistry area. Ishibashi and Imasaka et al. (13) have explored the use of the laser diode for molecular fluorometry, reporting it to be efficient, compact, rugged, and inexpensive. They have used an LD as a source for fluorometric enzymatic assay (14) and as a detector for HPLC determination of labeled protein (15). A frequency doubled LD was coupled to a fiber optic for use as an oxygen sensor based on fluorescence quenching of benzo[ghi]perylene (16). In this work, a study was made of various configurations of laser diodes and detectors for obtaining the lowest detection limits in practical and simple sample introduction schemes. In the first configuration, a cuvette was used to hold the sample. A comparison was made between a fluorescence spectrometer and a simple filter fluorometer. In another configuration, a flowing liquid jet fluorescence spectrometer was used. This configuration could be used as a flow injection or liquid chromatography detector. Finally, a compact cell glass filter fluorometer was constructed with minimal optics and cost.
EXPERIMENTAL SECTION Laser dye IR140 (Exciton Chemical Co., Inc., Dayton, OH; chemical name 5,5’-dichloro-ll-(diphenylamino)-3,3’-diethyl10,12-ethylenethiatricarbocyanine perchlorate) was dissolved in methanol and excited with a 200-mW continuous wave AlGaAs laser diode (SDL-2422-H1;Spectra Diode Labs, San Jose, CA). The laser wavelength was tuned to 794 nm by operating it at a temperature of -7.6 “C.Although the laser diode wavelength is in the near-IR region, the output could be seen as a weak red beam due to the high power of the device. This greatly facilitated alignment of the optics. A comparison was made of three different configurations of equipment used to focus the laser beam and measure fluorescence. In the cuvette fluorescence spectrometer system, the LD beam was processed with a collimating lens and an anamorphic prism pair (Melles Griot, Irvine, CA) to provide a collimated, squareshaped beam measuring 3 mm by 4 mm. This beam was passed through a methanol solution of the dye contained in a conventional 1cm X 1cm square cuvette. Fluorescence at 830 nm was collected at right angles with a 2 in. focal length lens and imaged on the entrance slit of a monochromator. The monochromator was a Spex Minimate 0.25 m (Spex Industries Inc., Edison, NJ) with a grating blazed for 750 nm and a red-sensitive (-2% quantum efficiency and -10 mA/W photocathodic sensitivity at 830 nm; spectral range 185 - 900 nm) photomultiplier tube (PMT) (HamamatsuR955, Hamamatsu Corp., Bridgewater, NJ) equipped with a thermoelectric cooler (Products for Research, Inc., Danvers, MA). For comparison, the monochromator (5-mm slits, spectral band-pass of 64 nm at 830 nm) was replaced by a Corion S10-830-F interference (10 nm fwhm and 45% transmittance at 830 nm) filter centered on the fluorescence peak. The fluorescence was excited as above but, this time, collected with the lens and sent directly 0 1989 American Chemical Society
862
ANALYTICAL CHEMISTRY, VOL. 61, NO. 8, APRIL 15. 1989
Table I. Analytical Figures of Merit for Different Instrumental Configurations fluorescence system
fg/mL
LOD mol/L
molecules
sensitivity, pA/(pg/mL)
liquid jet fluorescence spectrometer cuvette fluorescence spectrometer cuvette filter fluorometer compact cell glass filter fluorometer
1050 46 70 58000
1.3 X 5.9 x 1044 9.0 x 10-14 7.4 x 10-11
4.6 x 104 1.7 x 108 2.7 X lo8 3.4 x 10"
29 923 25500 16
blank noise, pA 50 70 3000
700
volume, p L 0.056 480 480 750
met e r
Monochromat o r
/
Figure 2. Schematic of fluorescence measurement in a liquid jet.
0
-250
0
+250
v e r t i c a l d i s t a n c e (urn) Figure 1. Spatial profile of LD beam focused by 20X microscope objective.
to the cooled PMT through the filter. The interference filter was used primarily to reject laser scatter. In the liquid jet fluorescence spectrometer system, the collimated, shaped laser diode beam was focused into a microscope objective using a 14 in. focal length lens. A 20X microscope objective focused the beam to a diameter of 450 pm. The diameter was measured by translating a 1O-pm pinhole along the focal plane. A Hamamatsu 32386-4513 photodiode (-0.6 A/W at 830 nm; spectral range 400-1100 nm) was used to monitor light flux passing through the hole. A plot of intensity vs pinhole position yielded a spatial profile of the focused laser (Figure 1). The sample was syringe pumped through Teflon tubing into a glass capillary with an internal diameter of 400 fim. The laser diode beam was focused into the liquid jet formed by the solvent immediately after exiting the capillary. Emission was collected by the lens, monochromator, and cooled PMT. A Corion RG850 colored-glass filter (long-pass, 50% transmittance at 850 nm; Corion Corp., Holliston, NJ) was used before the Spex Minimate 0.25-m monochromator to reject laser scatter (Figure 2). In the compact cell glass filter fluorometer system, the laser diode was placed as close as possible to the cuvette, and a lownoise, red-sensitive Hamamatsu S2386-45K photodiode was also placed in close proximity to the cuvette (see Figure 3). The colored glass filter was the only optical element used in this simple system as shown in Figure 3.
RESULTS AND DISCUSSION The laser diode is readily temperature tunable over a range of 20 nm; tunability is achieved with a thermoelectric cooler and thermistor built into the diode package. A built in photodiode monitors the output power and is used to stabilize the output. Cooling requires only a heat sink and a small fan. All laser diode fluorescence configurations resulted in high sensitivity although the systems with the cooled P M T were superior to the photodiode system. The best concentrational detection limit was achieved by passing a collimated laser diode beam through the cuvette. Replacing the monochomator with an interference band-pass filter increased the sensitivity of the measurements by a factor of 28, but cost a factor of 44 in laser scatter noise. The small monochromator also resulted in higher signal-to-noise (S/N) ratios than in-
Flgwe 3. Diagram of compact fluorometer: L, laser diode; s, sample; f, filter; d, photodiode; h, heat sink; p, power cord; b, bnc connector.
terference or glass filters, although a combination of filters may outperform the former. A comparison of the performance of the four fluorescence systems is given in Table I. In all cases, the limit of detection (LOD) was determined by noise on the scattered laser light reaching the detector. The liquid jet fluorescence spectrometer system gave the best absolute LOD, since this technique resulted in the smallest detection volume. The detection volume is described as a cylinder with diameter defined by the inside diameter of the capillary and length given by the width of the focused laser beam. The logarithmic calibration curve had a slope of 1.05 and was linear over 6 orders of magnitude; the sensitivity was 0.029 nA mL pg-'. The concentrational LOD was calculated by dividing the root mean square (RMS) noise on the blank by the sensitivity (linear calibration curve slope) and multiplying by 3. The RMS blank noise was estimated by dividing the peak to peak fluctuation by five. The IUPAC recommended value of k = 3 standard deviations of the RMS noise (17) was used to calculate the LOD. Based on the concentrational limit of detection and the volume of sample, the absolute LOD is 59 ag, or 46000 molecules, with a 1-s time constant of the measurement. With the liquid jet fluorescence spectrometer, a flow rate of 6.6 mL/min was used; the mass detection limit was calculated by the method of Dovichi et al. (10) mass detection limit = conc LOD X flow rate X time constant = (1.05 x 10-9 g/L)(i.ii x 10-4 L/s)(i s) = 1.2 x 10-13 and for a molecular weight of 779, this corresponded to 9 x lo7 molecules. The LOD was determined by the noise on the
Anal. Chem. 1989, 61, 863-871
laser scatter. Background from solvent contaminants was not observed, since there were few molecules that fluoresced in the red spectral region. Positioning the liquid jet at a 4 5 O angle with respect to the optical axis and parallel to the entrance slit greatly reduced the scatter reaching the monochromator entrance slit (18) although a higher sample flow rate was required to achieve laminar flow. The glass filter in front of the monochromator also reduced scatter, increasing the signal-to-noise ratio by a factor of 33. The compact cell glass filter fluorescence system gave poorer detection limits than the other three fluorescence setups discussed previously (see Table I). The instrument had an uncollimated laser beam for excitation and a photodiode for detection, achieving simplicity but a higher LOD. The uncollimated LD has a divergence angle of 35O by loo which resulted in inefficient irradiation of the detection volume, as well as greatly increased scatter. The lack of photodetector gain also limited the LOD of this system. The photodiode had a higher sensitivity at wavelengths greater than 900 nm compared to PMTs used in the other fluorescence spectrometer systems; this could have resulted in an interference from Raman scatter passed by the glass filter. The advantage of this instrument is its small size and simplicity. With a battery-operated laser diode, it could be completely portable. Three of the four laser diode fluorometer systems achieved concentrational detection limits of parts per trillion and below and the fluorescence system based on a liquid jet achieved an absolute detection limit of 46000 molecules. By use of a laser diode with higher irradiance and a sharp cut filter to minimize pickup of laser scatter, the LODs for all four fluorescence systems listed in Table I would be improved. In addition, focusing of the laser diode to a diameter of less than 450 I.cm would improve the LOD of the liquid jet system. Such work is now in progress.
863
ACKNOWLEDGMENT The authors gratefully acknowledge Rudy Strohschein of the Chemistry Department glassblowing shop for construction of the capillaries used in this work. Registry No. IR140, 53655-17-7.
LITERATURE CITED Hirshchfeld, Thomas Appl. Opt. 1978, 15, 2965-2966. Kirsch, Barbara; Volgtman, Edward; Winefordner, James D. Anal. Chem. 1985, 5 7 , 2007-2009. Mroz, Edmund A.; Lechene, Claude Anal. Eiochem. 1980, 102, 90-96. Haugen, (3. R.; Lytle, Fred E. Anal. Chem. 1981, 53, 1554-1559. Yamada, Sunao; Miyoshi, Fumihiro; Kano, Koji; Ogawa, Teiichiro Anal. Chlm. Acta 1981, 127, 195-198. Diebold, Gerald J.; &re, Richard N. Science 1977, 198, 1439-1441. Hershberger, L. W.; Callis, J. B.; Christian, G. D. Anal. Chem. 1979, 5 1 , 1444-1446. Mathies, Richard A.; Stryer, Lubert I n Applications of fluorescence in the Eiomedlcal Scbnces; Taylor, D. Lansing, Ed.; Alan R. Liss, Inc.: New York, 1986; Chapter 6. Kawabata, Yuji; Imasaka, Totaro; Ishibashi, Nobuhiko Talent8 1988, 33, 281-203. Dovichi, Norman J.; Martin, John C.; Jett, James H.; Trkula, Mitchell; Keller, Richard A. Anal. Chem. 1984, 56, 348-354. Nguyen, Dlnh. C.; Keller, Rlchard A.; Trkuia, Mbheli J. Opt. Soc. Am. E : Opt. Phys. 1987, 4 , 138-143. Johnson, Paul A.; Vera, Jorge A.; SmAh, Benjamin W.; Wlnefordner, James D. Spectres. Leff. 1988, 2 1 , 607-612. Imasaka, Totaro; Yoshltake, Akinori; Ishibashi, Nobuhiko Anal. Chem. 1984, 56, 1077-1079. Imasaka, Totaro; Okazakl, Takashi,; Ishibashi, Nobuhlko Anal. Chim. Act8 1988, 208, 325-329. Sanda, Kouji; Imasaka. Totaro; Ishibashi, Hobuhiko Anal. Chem. 1988, 58, 2649-2653. Okazaki, Takashi; Imasaka, Totaro; Ishibashi, Nobuhiko Anal. Chim Acta 1988, 209, 327-331. Long, Gary L.; Winefordner, James D. Anal. Chem. 1983, 55, 712A. Hirschy, Linda; Smith, Benjamin; Voigtman, Edward; Winefordner, James D. Anal. Chem 1982, 5 4 , 2307-2380.
.
RECEIVED for review September 27,1988. Accepted January 9, 1989. Research supported by DE-AS05-780R06022.
Simulation of Carbon- 13 Nuclear Magnetic Resonance Spectra of Alkyl-Substituted Cyclohexanones and Decalones G. Paul Sutton and Peter C. Jurs* Department of Chemistry, The Pennsylvania State University, 152 Davey Laboratory, University Park, Pennsylvania 16802
The carbon-13 nuclear magnetic resonance spectra for a serles of alkyl-substituted cyclohexanones and decalones have been simulated by uslng computer-aided methodology. Thls marks the first t h e that this software has been applled to carbonyl-contalnlng compounds. Multlple llnear regresslon analysis methods were used to create parametrlc equatlons relatlng observed chemlcal shlfts to a series of numerlcal descriptors encodlng structural features about each carbon center. The feaslbllity and usefulness of uslng alternative atom grouplng schemes prlor to model development were demonstrated, as dmerent model sets were used to assemble complete simulated spectra. The slmulated spectra were of sufflclent accuracy to allow dlstinctlon between palrs of geometrical Isomers. Partial slmulated spectra were also generated, such that the carbonyl carbon atom resonances were excluded, lndlcatlng the utility of creating and uslng slmulated subspectra. Slmulated spectra for an external predlctlon set of compounds were also generated and evaluated. 0003-2700/89/0361-0863$01.50/0
INTRODUCTION Carbon-13 nuclear magnetic resonance spectroscopy (13C NMR) is an important analytical tool that provides a wealth of structural information useful in the identification of organic compounds. The interpretation of complex 13C NMR spectra can often be aided by making computer-assisted spectral comparisons of observed spectra with library reference spectra; however, this approach is often limited due to lack of appropriate reference spectra. Spectrum simulation is one method by which chemists can generate approximate spectra that can be compared to the actual spectrum of an unknown. One approach to spectrum simulation involves the development of linear models that relate the observed chemical shift of a carbon atom to a series of numerically encoded structural parameters (descriptors). These models are of the form
S = bo + blXl
+ b2Xz + ... + bdXd
(1)
where S is the predicted chemical shift of a given carbon atom, X iis the value of the ith numerical descriptor, bi is the ith 0 1989 American Chemical Society
